Can the strength of a non-deductive argument be objectively evaluated?

Question

This question stems from an ongoing discussion I was having with @Bumble (link for anyone interested).

There are certain arguments that seem on the face of it very strong. For example, simply seeing my neighbor every day for 10,000 straight days may be used as a strong enough premise to conclude that I will see my neighbor tomorrow. On the other hand, “Charlie is a woman. Some women like poetry. Therefore, Charlie likes poetry.” may be seen as a weak argument. Yet in either case, the conclusion does not automatically follow from the premises.

The closest thing I can find that may objectify this process is the notion of logical probability. As quoted from the article,

Indeed, the logical interpretation, in its various guises, seeks to encapsulate in full generality the degree of support or confirmation that a piece of evidence e confers upon a given hypothesis h, which we may write as c(h,e)

After some responses to the logical interpretation, the article then introduces an evidential interpretation:

It may not be a matter of logic that the sun will probably rise tomorrow, given our evidence, yet there still seems to be an objective sense in which it probably will, given our evidence. In a crime investigation, there may be a fact of the matter of how strongly the available evidence supports the guilt of various suspects. This does not seem to be a matter of logic—nor of physics, nor of what anyone happens to think, nor of how the facts in the actual world turn out. It seems to be a matter, rather, of evidential probabilities.

But this notion is hotly contested as written in this snippet:

However, some authors are skeptical that there are such things as evidential probabilities—e.g. Joyce (2004). He also argues that there is more than one sense in which evidence tells for or against a hypothesis. Bacon (2014) allows that there are such things as evidential probabilities, but he argues that various puzzling results follow from Williamson’s account of them, in virtue of its identifying evidence with knowledge. Moreover, one may resist demands for an operational definition of evidential probabilities, while seeking some further understanding of them in terms of other theoretical concepts. For example, perhaps P(h∣e) is the subjective probability that a perfectly rational agent with evidence e would assign to h? Williamson argues against this proposal; Eder (2023) defends it, and she offers several ways of interpreting evidential probabilities in terms of ideal subjective probabilities. If some such way is tenable, evidential probabilities would presumably enjoy whatever applicability that such subjective probabilities have. This brings us to our next interpretation of probability.

The article then moves on to the notion of subjective probability but that itself is hotly contested in the remainder of that article.

This then begs the question: how does one objectively analyze the strength/weakness of a non deductive argument, or what it even means for an argument to be strong or weak? In the case of a deductive argument, we can at least seem to know objectively whether it is valid or invalid.

Answer 1

One of the problems with epistemic approaches to using probabilities, which includes Bayesianism, is that it treats propositions as isolated items. Each proposition is supposed to state an individual fact and the relations between propositions are expressed using conditional probabilities. It is a highly oversimplified model of how we reason about the real world.

Our broad knowledge of the world does not consist of propositions so much as entire theories or collections of theories. This is called epistemic holism and is an important part of Quine's philosophy. We hold a belief not because it can be calculated to have a high probability but because it is part and parcel of our best understanding of how the world works. Quine calls this our 'web of belief'. Some of our beliefs are central within the web. They are well-established and deeply entrenched and we are unwilling to revise them except in the face of extraordinarily good evidence to the contrary. Other beliefs are more peripheral and more readily revised. Our beliefs are held together by strands of deductive and non-deductive reasons.

Assessing the credibility of our beliefs is therefore not separable from the task of assessing what theories we consider to be the best. The criteria for a good theory include such things as: coherence, clarity, simplicity, comprehensiveness, empirical fit, lack of adhocness, consilience, explanatory value, practical value, etc. These criteria are not easily quantified and they may be in tension with one another. So different people may arrive at different conclusions about what are the best theories. That's just a fact of life. People disagree; even experts disagree.

But in practice there is a great deal of consensus about what our best theories are. Most physicists agree about most theories of physics. So we do not have a dichotomy of either everything is perfectly objective and computers could do all our thinking for us, or everything is completely subjective and there are no agreed criteria of rational thought. We are stuck in the middle trying to do our best.

Answer 2

Given your condition that there is no objective evaluation of the premises, I would be shocked if any objective conclusion can be reached... unless those premises turn out to be irrelevant to the conclusion. In this case even probabilistic evaluation doesn't help you because you are asserting that no objective estimate of probability is available.

A great deal of philosophy does rest upon subjective evaluation of assertions.

Those parts which don't, we currently tend to call scientific hypothesis.

Answer 3

How does one objectively analyze the strength/weakness of a non deductive argument, or what it even means for an argument to be strong or weak.

The purpose of what we call an "argument" is to make a particular statement more plausible to somebody, or, if possible, to convince somebody that something is the case, that a particular statement is true or a particular decision should be taken or is good to take. We can try to do so, using various means, some linguistic, some perhaps non-linguistic. Since the purpose of an argument is to either make sth more plausible or to convince someone, the "strength" of the argument could then be defined as the success rate in achieving that purpose. So, "the objective strength of an argument" -- that is, a given argument, a particular way of attempting to convince someone else (or oneself) or to make something more plausible and beliefworthy -- is how persuasive it actually is for particular people.

This can be studied either completely empirically -- from a psychological, sociological or historical point of view (e.g. Maya Bar-Hillel's exemplary studies of fallacies in probabilistic reasoning, which are on the border of psychology and philosophy) -- or with a more normative approach: when should we rationally or reasonably let ourselves be persuaded of something? One of the more normative philosophical approaches is for instance the way Jaakko Hintikka developed a logic of argumentation, explicitly conceptualizing the pragmatic, game-like dialogic context in which arguments take place. (See for instance, the SEP article about Logic and Games.) In this kind of approach, it's possible to define (and formalize) when a argument is "won".
(It's also interesting to see that the approach in the ancient Chinese Mohist canons was similar. See A.C. Graham's Later Mohist Logic, Ethics and Science.)

The discussion about evidential probabilities seems a rather niche, technical and theoretical discussion among some philosophers. It has no direct bearing on any particular evidence gathering to make a thesis more credible. No court cases will be suspended, no evidence will be found inadmissible (or admissible), either in science or in the court room, because those philosophers are not in total agreement about whether or not the concept of "evidential probability" is well-defined or applicable.

Answer 4

^{This answers an earlier version of the question.}

How can we objectively evaluate the strength of a non deductive argument? ... how does one objectively analyze the strength/weakness of a non deductive argument, or what it even means for an argument to be strong or weak?

I'm not sure where the assumption that an evaluation would be objective comes from.

An abductive conclusion needs to at least be consistent with the premises/observations. A conclusion inconsistent with the premises/observations can be ruled out.

Beyond that, assessment of an abductive argument is based on strength of evidence, prior beliefs, credibility and reliability of testimony, coherence, etc. There are many other Q&As on this site about those.

Abduction is "everyday reasoning."